Exploratory Data Analysis

load

library(tidyverse)
library(lubridate)
library(forecast)
library(ggrepel)
library(rvest)
library(plotly)
library(xtable)

dja <- read_csv("data/DJA.csv",skip = 4)
gold <- read_csv("data/GOLD_1791-2018.csv",skip = 2)
interest_rate <- read_csv("data/INTERESTRATE_1857-2018.csv",skip = 1)
sap <- read_csv("data/SAP_1871-2018.csv", skip=1)
cpi <- read_csv("data/USCPI_1774-2018.csv", skip=3)
gdp <- read_csv("data/USGDP_1790-2018.csv", skip=2)
wage <- read_csv("data/USWAGE_1774-2018.csv")
gdp_uk <-  read_csv("data/UKGDP_1700-2017.csv", skip=1)

brit_gold <- read_csv("data/GOLD_brit_1257-1945.csv", skip=1)
gold_london <- read_csv("data/GOLD_london_1718-2017.csv", skip = 1)

Down Jones Avg

el dja es una serie diaria, todas las demás son anuales.

dja <- dja %>% 
  mutate(Date = parse_date_time(Date,orders = "mdy"))

ggplotly(ggplot(dja,aes(Date, DJIA))+
  geom_line()) %>% 
  layout(legend = list(
      orientation = "h"))

summary(dja)

      Date                          DJIA         
 Min.   :1885-05-02 00:00:00   Min.   :   24.36  
 1st Qu.:1916-02-09 06:00:00   1st Qu.:   70.41  
 Median :1946-10-14 12:00:00   Median :  199.28  
 Mean   :1949-03-11 07:36:39   Mean   : 2354.03  
 3rd Qu.:1982-06-06 06:00:00   3rd Qu.: 1002.07  
 Max.   :2018-09-10 00:00:00   Max.   :26616.71  

dja %>% 
  mutate(dif = (DJIA - lag(DJIA, default = DJIA[1]))/lag(DJIA, default = DJIA[1])) %>% 
ggplot(.,aes(Date, dif))+
  geom_rect(fill="firebrick", 
            xmin=parse_date_time("01-01-1930",orders = "mdy"),
            xmax=parse_date_time("01-01-1940",orders = "mdy"),
            ymin=-1,
            ymax=1,
            alpha=0.5)+
    geom_line()

armo una lista de las crisis conocidas

Crisis

url <- "https://www.caproasia.com/2016/04/12/economic-crisis-since-1900-2015/"
crisis <- url %>%
  read_html() %>% 
  html_nodes(css = 'table') %>% 
  html_table(header = T)

crisis <- crisis[[1]] %>% 
  filter(Affected %in% c("United States","Global")) %>% 
  separate(Period,c("desde","hasta")," – ")

Expected 2 pieces. Missing pieces filled with `NA` in 10 rows [1, 4, 5, 6, 7, 8, 9, 10, 11, 12].

#en realidad las que terminan en "s" no duran toda la década. Las agrego a mano.
  # mutate(hasta = parse_date_time(case_when(grepl("s",desde)~as.numeric(str_extract(desde,"[[:digit:]]*"))+10,
  #                          TRUE~ as.numeric(hasta)),"y"),
  #        desde = parse_date_time(str_extract(desde,"[[:digit:]]*"),"y"))

crisis <- crisis %>% 
  mutate(hasta = parse_date_time(case_when(desde=="1970s"~"1979",
                           desde=="1980s"~"1982",
                           desde == "1990s"~"1991",
                           TRUE~hasta),"y"),
         desde = parse_date_time(case_when(desde=="1970s"~"1973",
                           desde=="1980s"~"1981",
                           desde=="1990s"~"1990",
                           TRUE~desde),"y"))

# tabla <- crisis %>%
#   filter(Affected %in% c("United States","Global")) %>%
#   select(-Region, - Affected) %>%
#   xtable(.)

#En la consola
# print(tabla, include.rownames = F)

crisis_largas <- na.omit(crisis)
crisis_puntuales <- crisis %>% 
  filter(is.na(hasta))


dja <- dja %>% 
  mutate(dif = (DJIA - lag(DJIA, default = DJIA[1]))/lag(DJIA, default = DJIA[1])) 
ggplot()+
  geom_rect(data= crisis_largas, 
            aes(xmin=crisis_largas$desde,
            xmax=crisis_largas$hasta),
            fill="firebrick", 
            ymin=-1,
            ymax=1,
            alpha=0.5)+
    geom_line(data = dja,aes(Date, dif))+
  geom_vline(data=crisis_puntuales, aes(xintercept=desde), color = "red", linetype="dashed")

NA
NA

Este gráfico me da la sensación de que todo estuviera corrido a la derecha (mirando las crisis puntuales vs los picos)

Gold

gold %>% 
  ggplot(., aes(Year, `New York Market Price (U.S. dollars per fine ounce)`))+
  geom_line(size=1)+
  geom_vline(xintercept = 1971, color = "red")+
  geom_label_repel(data=data_frame(),aes(x=1971,y=1000,label="Fin del patrón oro"),nudge_x = -5,force=10,size=7)+
  theme_minimal()+
  labs(y="Dólares por onza de oro", x="Año", title="Precio Oro Mercado de Nueva York",
       subtitle = "Precio por onza. 1790-2017")+
  theme(text = element_text(size = 20))


ggsave("plots/oro.png", dpi=300, width = 10, height = 7, scale=1)

gold %>% 
  ggplot(., aes(Year, `New York Market Price (U.S. dollars per fine ounce)`))+
  geom_line(size=1)+
  geom_vline(xintercept = 1971, color = "red")+
  geom_label_repel(data=data_frame(),aes(x=1971,y=1000,label="End of the gold standard"),nudge_x = -5,force=10,size=7)+
  theme_minimal()+
  labs(y="dollars per fine ounce", x="Year")+
  theme(text = element_text(size = 20))

ggsave("plots/oro_en.png", dpi=300, width = 10, height = 7, scale=1)

interest_rate

graf <- interest_rate %>% 
  gather(type,rate,2:4) %>% 
  ggplot(., aes(Year,rate,color=type))+
  geom_line()+
  guides(color=guide_legend(nrow=2,byrow=TRUE))+
  theme(legend.position = "bottom")
ggplotly(graf) %>%
  layout(legend = list(
      orientation = "h"
    )
  )

• La tasa de largo plazo es una serie mucho más suave (eso es un dato conocido de finanzas no?)
• Los surplus funds también parecen ser más volátiles hasta los 40

sap

sap %>% 
  summary()
sap %>% 
  gather(type, value,2:4) %>%
  mutate(type= case_when(type=="The S&P Index Average for January"~"The S&P Index\nAverage for January",
                         type=="The Accumulated S&P Index Average for January"~"The Accumulated S&P\nIndex Average for January",
                         TRUE~type)) %>% 
  ggplot(.,aes(Year,value, color=type))+
  geom_line()+
  facet_grid(type~.,scale="free")+
  theme(legend.position = "bottom",
        strip.text.y = element_text(angle = 0))

ts(sap$`Annual Yield`, start=min(sap$Year), frequency = 1) %>% 
  na.omit() %>%
  auto.arima(.)

CPI

cpi %>% 
  ggplot(., aes(Year, `U.S. Consumer Price Index *`))+
  geom_line(size=1)+
  geom_vline(xintercept = 1971, color = "red")+
  geom_label_repel(data=data_frame(),aes(x=1971,y=100,label="Fin del patrón oro"),nudge_x = -5,force=10,size=7)+
  theme_minimal()+
  labs(y="IPC", x="Año")+
  theme(text = element_text(size = 20))

ggsave("plots/cpi_orig.png", scale = 1)

gdp

me interesa ver el PBI normalizado por el crecimiento poblacional, y además normalizado por la cantidad de oro que puede comprar (en lugar de normalizar por el CPI):

gdp <- left_join(gold, gdp, by = "Year") %>% 
  mutate(gdp_in_gold = `Nominal GDP per capita (current dollars)`/`New York Market Price (U.S. dollars per fine ounce)`,
         Year = parse_date_time(Year,"y")) 
  
ggplotly(ggplot(gdp,aes(Year,gdp_in_gold))+
  geom_line())

A partir del 1900 pareciera que se arman 3 ciclos muy largos

• 1914-1933
• 1933-1980
• 1980-2012

Agregando referencias históricas de las crisis conocidas # gdp_in_gold_eda.PNG

ggplot()+
geom_rect(data= crisis_largas, 
            aes(xmin=crisis_largas$desde,
            xmax=crisis_largas$hasta),
            fill="firebrick", 
            ymin=-Inf,
            ymax=Inf,
            alpha=0.5)+
  geom_line(size=1,
    data = gdp %>% 
              filter(Year>parse_date_time(1900,"y"))
            ,aes(Year, gdp_in_gold))+
  geom_vline(data=crisis_puntuales, aes(xintercept=desde), color = "red", linetype="dashed")+
  geom_vline(xintercept = parse_date_time(1971,"y"),color = "gold")+ #fin del patron oro
  # scale_x_datetime(date_breaks = "15 years",labels = date_format("%Y") )+
  scale_x_datetime(breaks = parse_date_time(seq(1900,2017, 15),'y'), labels = date_format("%Y"))+
  theme_minimal()+
  labs(x="", y="PBI en oro", title="PBI Estados Unidos",
       subtitle= "Millones de onzas de oro, 1900-2017")+
  theme(text = element_text(size = 20))

Error in FUN(X[[i]], ...) : object 'gdp_in_gold' not found

La guerra de sesesión de EEUU fué entre el 12 de abril de 1861 y el 9 de abril de 1865

A partir de ahí el pbi en oro crece hasta el fin del patron oro.

GDP complementario

wage

Podemos deflactar el salario horario por el CPI

ggplotly(
wage %>% 
  left_join(cpi,by="Year") %>% 
  na.omit() %>% 
  mutate(salario_horario_real = `Production Workers Hourly Compensation (nominal dollars)`/`U.S. Consumer Price Index *`) %>% 
  ggplot(.,aes(Year,salario_horario_real))+
  geom_line() 
  )

wage %>% 
  summary()

Podemos deflactar el salario horario por el CPI

ggplotly(
wage %>% 
  left_join(cpi,by="Year") %>% 
  na.omit() %>% 
  mutate(salario_horario_real = `Production Workers Hourly Compensation (nominal dollars)`/`U.S. Consumer Price Index *`) %>% 
  ggplot(.,aes(Year,salario_horario_real))+
  geom_line() 
  )

wg_gold <- wage %>% 
  filter(Year>=1900) %>% 
  left_join(gold, gdp, by = "Year") %>% 
  mutate(wg_in_gold = `Production Workers Hourly Compensation (nominal dollars)`/`New York Market Price (U.S. dollars per fine ounce)`,
         Year = parse_date_time(Year,"y")) %>% 
  na.omit()  


ggplot()+
geom_rect(data= crisis_largas, 
            aes(xmin=crisis_largas$desde,
            xmax=crisis_largas$hasta),
            fill="firebrick", 
            ymin=-Inf,
            ymax=Inf,
            alpha=0.5)+
  geom_line(size=1,data = wg_gold, aes(Year, wg_in_gold))+
  geom_vline(data=crisis_puntuales, aes(xintercept=desde), color = "red", linetype="dashed")+
  geom_vline(xintercept = parse_date_time(1971,"y"),color = "gold")+ #fin del patron oro
  scale_x_datetime(breaks = parse_date_time(seq(1900,2017, 15),'y'), labels = date_format("%Y"))+
  # scale_x_datetime(date_breaks = "15 years",labels = date_format("%Y") )+
  theme_minimal()+  
  labs(x="", y="Salario en oro", title="Salario horario Estados Unidos",
       subtitle="Onzas de oro, 1900-2017")+
  theme(text = element_text(size = 20))


ggsave("plots/wg_in_gold_eda.PNG", dpi = 300, width = 10,height = 6)

ggplot()+
geom_rect(data= crisis_largas, 
            aes(xmin=crisis_largas$desde,
            xmax=crisis_largas$hasta),
            fill="firebrick", 
            ymin=-Inf,
            ymax=Inf,
            alpha=0.5)+
  geom_line(size=1,data = wg_gold, aes(Year, wg_in_gold))+
  geom_vline(data=crisis_puntuales, aes(xintercept=desde), color = "red", linetype="dashed")+
  geom_vline(xintercept = parse_date_time(1971,"y"),color = "gold")+ #fin del patron oro
  #scale_x_datetime(date_breaks = "15 years",labels = date_format("%Y") )+
  scale_x_datetime(breaks = parse_date_time(seq(1900,2017, 15),'y'), labels = date_format("%Y"))+
  theme_minimal()+  
  labs(x="", y="Wage in gold")+
  theme(text = element_text(size = 20))


ggsave("plots/wg_in_gold_eda_en.PNG", dpi = 300, width = 10,height = 6)

Se ven los mismos tres períodos. Pero a diferencia del GDP, el período 1980-2012 tiene un nivel más bajo que el anterior.

¿ Si quisieramos comparar ingrsos con algún revenue tendríamos usar S&P o DJA?

Complementario

UK

oro <- full_join(brit_gold,gold_london)

Joining, by = "Year"

names(oro)

[1] "Year"                                                                                  
[2] "British Official Price (British pounds per fine ounce end of year)"                    
[3] "London Market Price (British £ [1718-1949] or U.S. $ [1950-2011] per fine ounce)"

#de http://fx.sauder.ubc.ca/data.html
tc <- read_csv("data/ex_rate.csv")

Parsed with column specification:
cols(
  `MMM YYYY` = col_character(),
  `GBP/USD` = col_double()
)

library(lubridate)

tc <- tc %>% mutate(date = parse_date_time(`MMM YYYY`,"my"),
              Year = year(date)) %>% 
  group_by(Year) %>% 
  summarise(gbp_usd = mean(`GBP/USD`))


#de http://fx.sauder.ubc.ca/etc/USDpages.pdf

tc_1950_1970 <- data_frame(Year=1950:1970, gbp_usd = 0.35714) %>% 
  mutate(gbp_usd = case_when(Year ==1967 ~ 0.36210,
                             Year >1967 ~ 0.41667,
                             TRUE ~ gbp_usd))

`data_frame()` is deprecated, use `tibble()`.
This warning is displayed once per session.

tc <- bind_rows(tc_1950_1970,tc)
tc

expreso al oro siempre en pounds

tail(oro)


oro$`London Market Price (British £ [1718-1949] or U.S. $ [1950-2011] per fine ounce)`

oro <- oro %>% 
  filter(Year %in% c(1700:2017)) %>% 
  mutate(serie_unificada = case_when(Year < 1718 ~ `British Official Price (British pounds per fine ounce end of year)`,
                                     Year >=1718 ~`London Market Price (British &pound; [1718-1949] or U.S. $ [1950-2011] per fine ounce)`))

## Tengo que pasar todo a libras, desde 1950 al serie esta en dólares

oro <- oro %>% 
  left_join(tc) %>% 
  mutate(serie_unificada = case_when(Year>1950 ~ serie_unificada*gbp_usd,
                                     TRUE ~ serie_unificada))
  

Gráfico Oro

  oro %>% 
  filter(Year<=1900) %>% 
  ggplot(., aes(Year,serie_unificada))+
  geom_line(size=1)+
  scale_y_continuous(limits = c(2,6))+
  theme_minimal()+
  labs(y="British pounds per fine ounce", x="Year")+
  theme(text = element_text(size = 20))


ggsave("plots/oro_uk.png", dpi=300, width = 10, height = 7, scale=1)

PBI uk en oro

gdp_uk <- gdp_uk %>% 
  left_join(oro) %>% 
  mutate(gdp_in_gold = `Nominal GDP (million of pounds)`/serie_unificada)

Joining, by = "Year"

crisis <- gdp_uk %>% 
  filter(Year %in% c(1700:1900)) %>% 
  mutate(crisis =gdp_in_gold) %>% 
  filter(Year %in% c(1794,1803, 1812, 1822,1833,1842,1850, 1858, 1868,1879, 1885, 1893))

gdp_uk %>% 
  filter(Year %in% c(1700:1900)) %>% 
ggplot(., aes(Year, gdp_in_gold))+
  geom_line(size=1)+
  geom_text_repel(data = crisis, aes(Year, crisis,label=Year),nudge_x = 5, nudge_y = -20,force=12,size=4)+
  theme_minimal()+
  labs(y="PBI en oro", x="Año", title= "Producto Bruto Interno Reino Unido", 
       subtitle="Millones de onzas de oro. 1700-1900")+
  theme(text = element_text(size = 20))


ggsave("plots/uk_gdp.png",scale = 1)

Saving 7.29 x 4.5 in image

gdp_uk %>% 
  filter(Year %in% c(1700:1900)) %>% 
ggplot(., aes(Year, gdp_in_gold))+
  geom_line(size=1)+
  geom_text_repel(data = crisis, aes(Year, crisis,label=Year),nudge_x = 5, nudge_y = -20,force=12,size=4)+
  theme_minimal()+
  labs(y="GDP in gold", x="")+
  theme(text = element_text(size = 20))

Error in FUN(X[[i]], ...) : object 'gdp_in_gold' not found

gdp_uk %>% 
  write_csv("data/gdp_uk_gold.csv")

complemento UK

ggplot()+
geom_rect(data= crisis_largas, 
            aes(xmin=crisis_largas$desde,
            xmax=crisis_largas$hasta),
            fill="firebrick", 
            ymin=-Inf,
            ymax=Inf,
            alpha=0.5)+
  geom_line(size=1,
    data = gdp_uk %>%
      mutate(Year=parse_date_time(Year,"y")) %>% 
              filter(Year>parse_date_time(1900,"y"))
            ,aes(Year, gdp_in_gold))+
  geom_vline(data=crisis_puntuales, aes(xintercept=desde), color = "red", linetype="dashed")+
  geom_vline(xintercept = parse_date_time(1971,"y"),color = "gold")+ 
  scale_x_datetime(breaks = parse_date_time(seq(1900,2017, 15),'y'), labels = date_format("%Y"))+
  # scale_x_datetime(date_breaks = "15 years",labels = date_format("%Y") )+  
  theme_minimal()+
  labs(x="", y="GDP in gold")+
  theme(text = element_text(size = 20))

ggsave("plots/gdp_uk_complement.png", dpi=300, width = 10, height = 7, scale=1)

PBI mundial